-
Progress In Electromagnetics Research M, Vol. 45, 195–207,
2016
Torque and Ripple Improving of a SR Motor Using Robust
ParticleSwarm Optimization of Drive Current and Dimension
Abbas Ketabi, Ata Yadghar*, and Mohammad J. Navardi
Abstract—In this paper, the robust optimization shape and drive
of switched reluctance motors(SRMs) are discussed using robust
particle swarm optimization (RPSO). The shape optimum goalof the
algorithm was found for maximum torque value and minimum torque
ripple, following changingthe geometric parameters. The drive
optimum aim of the algorithm was found minimum torque
ripple,following changing the current profiles. The optimization
process was carried out using a combinationof RPSO and Finite
Element Method (FEM). Fitness value was calculated by FEM analysis
usingCOMSOL 4.2, and the RPSO was realized by MATLAB 2011. The
proposed method has been appliedto two case studies and also
compared with seeker optimization algorithm. The results show
thatthe optimized SRM using RPSO has higher torque value, lower
torque ripple and higher robustness,indicating the validity of this
methodology for SRM design and implementation.
1. INTRODUCTION
Switched reluctance motors (SRMs) are widely used in different
applications such as new generationof transportation. Simple and
strong structure, high efficiency, low weight and small size due to
theabsence of rotor windings, variable speed operations, high
torque to inertia ratio, reliability, relativelylow manufacturing
costs and fault permissibility of SRM are reasons for choosing a
SRM in industryespecially in electrical vehicles [1]. High starting
torque for initial acceleration, high average torque,and high
efficiency to save battery for a longer operating time are the main
factors to design such apractical motor [2]. Furthermore, these
features make SRM a viable alternative to other commonly usedmotors
such as AC, BLDC, PM Synchronous or universal motors for numerous
applications. The mostimportant disadvantage of SRM is its high
torque ripple in comparison with other electrical motors.It causes
acoustic noises and mechanical vibration. Applying discrete current
to sequenced phase isthe main reason of torque pulsation. There are
two known methods for torque ripple minimization: I)driver
optimization, II) geometry change [2]. The geometry of a SRM
consists in different parametersthat affect the features of the
motor. Various researches on geometries and drive design of SRM
andgenerators have been successfully investigated in the past. In
[1], a FEM analysis showed that the effectsof phase current
chopping of a sample 8/6 SRM could reduce the core loss about 50%.
Nevertheless, noexact explanation about the frequency and amplitude
was provided. A torque sharing function theorywas introduced in
[2]. In [3], the phase current profile optimized and compared to
constant phase currentprofile. In [4], the changing of EMF was used
for modifying torque, where the phase current overlappingwas
integrated to reduce torque ripple where the motor simulated with
mathematical equation and FEManalysis was ignored. In [5], a PI
filter was employed to create novel phase voltages, to obtain a
ripplefree torque by a PWM digital signal processor device. Ref.
[6] compared a PWM voltage generator oncewithout commutation and
once again with an optimized phase commutation angle and concluded
thatthe commutation angle method revealed the efficiency. Using PWM
direct torque observation during
Received 22 November 2015, Accepted 29 December 2015, Scheduled
12 January 2016* Corresponding author: Ata Yadghar
([email protected]).The authors are with the University
of Kashan, Iran.
-
196 Ketabi, Yadghar, and Navardi
the movement of the rotor is a method proposed in [7].
Unfortunately, direct control of torque is notconfirmable as the
best method of motor features optimization. In [8], by NSGA-II
optimization methoddecided the constants in PI-Filter of the PWM
and also the values of cut-off and cut-on angles whencommuted
between phases as in [9] which suggested two values for the end of
raising and the beginning offalling current in each phase. In [10],
a new method of torque sharing function was introduced. Insteadof
bipolar excitation of a SRM, in [11] a single pole excitation was
proposed, which can reduce the torqueripple but increase the price
of product. In [12], a fitness function by weighting two objective
functions(torque ripple and copper loss) was proposed but without
introducing the optimization method. In [13],besides the speed
stability of a BLDC motor that worked similarly to a SRM, it was
claimed that thedesigned drive was able to correct the power factor
by the advantages of using isolated-zeta converter.In such a
context, many researches have successfully addressed the drive
optimization problem [14–17].
Seeker Optimization Algorithm (SOA) has been used in SRM
geometric optimization [18]. Themost efficient aim of the algorithm
was found for maximum torque value at a minimum mass of theentire
construction, following changing the geometric parameters. Normal
back-EMF method wasreplaced by an integral variable structure
module in [19, 20] and operated in three options, two
phaseconduction mode, low speed commutation mode, high speed
commutation mode for different cases ofmotor performance. In [21,
22] besides the consideration of core saturation of the SRM, a
selectionmethod of reference current points from outgoing to
incoming phase is introduced to optimize thetorque ripple. DE
optimization has been applied to design a low torque ripple and
copper loss and ahigh average torque motor by modifying the angle
and dimension of stator and rotor poles [23, 24].
Until recently, the robust design issues in the optimal geometry
SRM design have been barelyaddressed [25]. From a mathematical
point of view, SRM design problem can be modeled as anoptimization
problem and be solved, but in the implementation, manufacturing
tolerances of motorparameters are challenge. Therefore,
optimization methods are required that can deal with
unavoidableuncertainties (noise factors) in the industrial
manufacturing process, such as material characteristicsand
manufacturing precision [26]. The practice of optimization that
accounts for uncertainties andnoise is often referred to as robust
optimization [27–30]. In this paper, a new method of optimization
isemployed to find a practical and ready to manufacture design of
drive and geometry of SRM. The basealgorithm of optimization is
originated from Particle Swarm Optimization (PSO) method
integratedwith a neighborhood looking-up lemma to provide the
robustness of the design. This feature of thedesign permits a range
of parameters, to prevent the unwelcome errors effects on design
objectivefunctions. The aim of our investigation is to reach
maximum torque value at a minimum torque ripple.To get the optimum
shape and drive performance of the motor, a combination of FEM and
Robust PSO(RPSO) is organized. The fitness value of the objective
functions is estimated by COMSOL 4.2, and theoptimization
individuals are decided using MATLAB by the applied RPSO algorithm.
The proposedmethod has been applied to a case study, and it has
also been compared with Seeker OptimizationAlgorithm [18].
This paper is organized as follows. Section 2 describes the
principle of SRM and drive performance.Section 3 proposes a RPSO
method for designing a current profile of drive and also the
geometry ofrotor and stator. Section 4 discusses the results, and
finally the conclusion is drawn in Section 5.
2. PRINCIPLES OF SWITCHED RELUCTANCE DESIGN
2.1. SRM Geometry Design Characteristics
SRM have salient poles that the windings are concentrated on
stator poles. The phase induction varies,since the related air gap
between rotor and stator poles changes in every other period [20].
The procedureof rotation energizes the stator phase and pulls the
rotor pole towards the related stator pole. Thereare different
types of winding excitation, but in our case of study 6/4 SRM no
more than two phasesare conducted. During the excitation phase
voltage equation is derived as:
v = R · i + ∂λ∂t
(1)
where, v, i and λ are the phase voltage, phase current and flux
linkage. Determination of the numbersof stator phases and rotor
poles is very discussable since the higher number of poles
decreases the torque
-
Progress In Electromagnetics Research M, Vol. 45, 2016 197
ripple and fewer poles help reducing the loss of switching. In
this paper a 6/4 SRM has been chosendue to its modified trade-off
between loss and torque ripple. Thanks to previous researches, a
6/4 SRMprovides a wider range of constant power and torque
production than its low switching loss [18]. Theco-energy of a SRM
is calculated as:
W ′f =12L(θ) · i2 (2)
where L is the inductance of the air gap field changed through
the changing of θ, and i is the phasecurrent which is variable in
this paper due to the modification of the phase current profile.
Then theelectromagnetic torque is given by partial derivative of
the co energy with respect to theta.
Tf =∂W ′f∂θ
=12i2 · ∂L(θ)
∂θ(3)
The induction varies during the movement of rotor as:
L(θ) =N2
R(θ) (4)
where N is the number of turns per phase and R the reluctance
between rotor and stator, which is afunction of θ.
As well known, the torque ripple and torque output are sensitive
to geometry variables of statorand rotor, and their selection is a
vital part of SRM design process [18]. Therefore, for the
shapeoptimization, six design parameters, shaft radius, rotor pole
angle, rotor inner radius, rotor outerradius, stator inner radius
and stator pole angle, as shown in Fig. 1, are selected as design
variables.Optimization process is programmed to maximize the
average torque and to minimize the torque rippleby defining the
following equations:
Tave =1Θ
Θ∫
0
T (θ)dθ (5)
Torque Ripple =Tmax − Tmin
Tave(6)
where T (θ) is electromagnetic torque at the rotor position θ, Θ
the associated electrical period equal toninety degrees, and Tmax
and Tmin are maximum and minimum of T (θ) over one period Θ,
respectively.
The objective function, which will satisfy both conditions of
maximum average torque and minimumripple torque, is a fitness
function defined as in Eq. (7).
fitness function =Torque Ripple
Tave(7)
In this article, the electromagnetic torque per meter of axial
thickness has been computed by a two-dimensional finite element
method (2-D FEM). Since the torque is linearly proportional to the
motoraxial length, the calculated torque per meter needs to be
multiplied by the motor axial length. COMSOL4.2 meshes the rotor
and stator poles and the space between them in triangular form and
then solvesstationary Maxwell’s equations by a linear or nonlinear
solver automatically. The software calculatesthe co-energy and
torque through air gaps in every 2-degree rotation of the rotor.
All the results aresaved in an array in MATLAB 2011. For SRM, there
is an optimum angle called “firing angle” betweentwo stator poles,
which the controller should switch between windings and turn on the
next phase forbuilding a rotating magnetic field [18]. The firing
angle (FA) equation is as following.
This research is highly concerned about the torque average and
torque ripple of the motor, so infitness function the mass of the
motor has been ignored.
2.2. SRM Drive Characteristics
Switched reluctance machine is a cost effective solution in
rotary applications. Besides the amount ofmaterials used in a SRM,
the cost of power electronics converter should be considered. Using
DSPdevices, the production of determined phase current in the rated
speed becomes easier [6]. Also it
-
198 Ketabi, Yadghar, and Navardi
Figure 1. Optimization variables.
provides independent control of individual phases. Providing
accurate and pre-determined currentrequires 3 current source
inverters in the case of utilizing a 6/4 SRM. The drive is
commanded by apre-determined current curve that is optimized. The
value of phase current can be varied in every2-degree rotation
during the rotor movement. Fig. 2 shows a microprocessor based
drive, used to ignitethe transistor in each phase.
Figure 2. SRM typical drive.
In this case, 3 of these drives have been employed. The duty
cycle of the buck DC-DC convertordetermines the value of falling
current. In [18], the DC current source set to the maximum current
neededfor phase excitation. There are several ways to minimize the
torque ripple using drive performance, suchas 1) phase overlapping
and 2) hysteresis band setting [14]. But a less investigated item
is phase currentprofiling. Fig. 3 shows the ideal current
commutation and profile for a three-phase motor. Each phaseis
applied to the related current by 30 degrees, and the incoming
phase rises right after the outgoingphase is turned off. The
objective of obtaining a new current phase profile is to reach a
flat and freetorque ripple [9]. A fast and also cheap device is
needed to change the dc current source by every 2degrees of rotor
rotation. Therefore, for the drive optimization, 15 design
parameters, phase currentin every 2 degrees over one 30 degrees,
are selected as design variables. It has pre-programmed bythe
optimization output and influenced by the position of rotor and its
polar distance to aligned andunaligned positions.
The objective function for the drive optimization problem, which
will satisfy both conditions ofmaximum average torque and minimum
ripple torque, is a fitness function defined as in Eq. (7).
There are several advantages for pre-programmed current
controller. It can decrease the cost ofproduction, since the torque
feedback devices can be neglected. Also simple robust current can
be
-
Progress In Electromagnetics Research M, Vol. 45, 2016 199
Figure 3. Ideal phase communication of SRM.
produced without the use of compensator lead networks.
Transistor fails because overcurrent comingfrom the source can be
prevented. Also by a pre-design solution, the saturation of the
motor excited ironcan be avoided. This research has tried to
achieve an appropriate profile of torque with pre-determinedfall of
current in special position of the rotor during the excitation of
each phase. The overlapping ofincoming and outgoing phases has been
neglected according to the ideal current commutation.
2.3. Robust Design
From a mathematical point of view, SRM design problem can be
modeled as an optimization problemand be solved, but in the
implementation, manufacturing tolerances of motor parameters and
accuracytolerance of motor drive current are design challenges.
Hence, for real-world scenarios, optimizationmethods are needed,
which can deal with these uncertainties, and solutions ought to be
found, whichare not only optimal in the theoretical sense, but also
practical in real life. The practice of optimizationthat accounts
for uncertainties and noise is often referred to as robust
optimization. Here “robust” isintended in the sense that the
optimum found by the algorithm is not too sensitive with respect
tosmall changes in the parameters. In this article, the robust
optimum with respect to manufacturingtolerances and drive current
tolerance is obtained by the application of the RPSO algorithm.
3. ROBUST PARTICLE SWARM OPTIMIZATION
3.1. Particle Swarm Optimization
In past several years, PSO has been successfully applied in many
researches to find an optimised solutionfor a problem. Many
researchers believe that results are obtained faster and more
accurate than othermethods [26]. PSO is a population based
stochastic optimization technique [26]. Such as GeneticAlgorithms,
PSO stocks many similar results with evolutionary computation
techniques. Like manyevolutionary algorithms, the system begins
with an initial stochastic population and continues
withoptimization of the generations. PSO does not have mutation and
crossover unlike GA. There arepotential solution individuals in
PSO, which fly all over the problem by means of defined vector.
Theparticles move through a defining space by the guidance of best
solutions called “PBEST”. Anotherbest solution is obtained by the
competition between neighborhoods of the best fitness value
individualscalled “GBEST”. According to the position of the bests,
PSO gives a variable velocity in each step tothe solutions. Few
parameters should be set before solving a problem. It is one of the
reasons that PSOis applied to many researches and becomes more
attractive in recent years.
3.2. RPSO
Normally in common optimization, finding the optimum of the
objective function is the only task thatthe operator expects to
receive [28]. In real world, there are multiple factors affecting
the optimizationprocess. In general, there are three types of
uncertainties approaches for practical design as shown inFig. 5
[28–30].
-
200 Ketabi, Yadghar, and Navardi
Figure 4. RPSO algorithm.
I) Deterministic: Providing a range of Pareto front and letting
the client responsible for choosing asolution.
II) Reliability: if deterministic solutions limited to a range
of smoother Pareto front the client canchoose its desired solution
from a wider range with low tolerance of objectives.
III) Robust: There are several situations, in which a scientist
decides to solve the optimizationproblem by robust optimization.
First the inappropriate conditions of environment and
operationalcircumstances. Second, the parameters might change after
the motor operation or during theprocedure of manufacturing. Third,
the system itself may produce noisy outputs. So seeking for asmooth
and flat area in a Pareto curve is the research goal, called robust
optimization.
Fig. 5 also declares valleys in Pareto front that causes
problems in the case of not accurate cuttingof motor iron or shape
changes due to an accident during operation. In the same figure, a
hatched
-
Progress In Electromagnetics Research M, Vol. 45, 2016 201
(a) (b) (c)
Figure 5. Comparison among reliable, deterministic and robust
selections of Pareto front candidates.
appropriate areas can be seen for robust optimization [29]. In
general, the robust optimization mainequation can be defined as
follows:
fexp =
+∞∫
−∞f(x + δ) · pdf(δ)dδ (8)
where δ is the distributed according to the probability function
pdf(δ). So robust optimization solvedresult can compete with its
neighbors, and also the objects in the neighborhood are so close to
the bestsolution. In this algorithm, there are three vectors
searching for all individuals in each N -dimensionalenvironment,
where N is the dimension of search space. A major impact of
comportment of RPSO isits neighborhood topology. Particles are
connected to other particles and affected by the neighborhoodcalled
the neighborhood topology. The procedure of optimization is similar
to the PSO main algorithm,but there are some changes in vector
directions of movement through the N -dimension space and alsothe
velocity. A brief statement is given here about how our RPSO works.
There is no better way forinitialization of a basement of
individuals for being optimized except the random initial
population.Once the initial population established, the velocity
starts to change by the current position of PBEST.In the algorithm
as PSO, PBEST is the representative of the objective function value
of the best positionof the particle. The current position and
velocity are established randomly for the first iteration. Thenthe
particles are updated by so-called change rule [31]:
vi = (vi + U(0, ϕ) ∗ (pi − xi) + U(0, ϕ) ∗ (Pgnbh − xi) (9)where
U(0, ϕ) is an N -dimensional vector that is uniformly distributed
between 0 and ϕi, and gnbhrepresents the global neighborhood of
particle I according to Fig. 6. Then the position updated as
thefollowing equation [32–34];
Current Position(xi+1) = Current Position(xi) + vi (10)
This algorithm updates all particles’ personal bests in the
first step, then moves the particles accordingto the defined
velocity. As normal PSO, a stop criterion is used to exit the loop.
Also another optionin the algorithm is used to not stop until a
satisfying result is obtained. The overview concept ofRPSO
algorithm is defined in Fig. 4 according to [27] including a
modification algorithm in changingthe velocity vector for
considering the location of neighbors.
4. RESULTS AND DISCUSSION
In this research, the RPSO algorithm has been employed to solve
two independent optimization problems1) SRM geometry robust design
and 2) SRM drive robust design using MATLAB software. On theother
side a MATLAB linked Finite Element Software, COMSOL 4.2 is used to
torque calculation andobjective functions evaluation. COMSOL 4.2
contains a wide range of geometries that are helpful
forestablishment different types of motor containing SRM. Besides,
it provides an online link to MATLAB
-
202 Ketabi, Yadghar, and Navardi
that exchanges the data during the optimization of population
and generations. All the executedalgorithms run on a PC with CORE
i7 processor and 8 GB RAM.
This section has two parts. In the first part, the results of
SRM geometry robust design usingRPSO is compared with previous work
that has optimized the same SRM structure using the
SeekerOptimization Algorithm [18].
In the second part, the results of SRM drive robust design using
RPSO is presented and comparedwith the results of setting the
current of each phase to a constant value equal to the maximum
currentwhere the geometry parameters are according to the results
of Seeker Optimization Algorithm [18].
4.1. SRM Geometry Robust Design
The optimized variables of SRM geometry is expressed in Table 1.
Fortunately, the RPSO model weightis equal to the weight of the
presented model in [18]. Both of them are 0.63 kg in each cm of
depth. Itshows that the saturation of the core is not a concern
anymore because the amount of air is not lessthan the reference
motor and the reluctance doesn’t change.
Table 1. Geometry parameters of optimised result and the worst
random neighbor and the deviationsof worst neighbor from the main
result.
Geo
met
ry
Mod
el
Shaf
t R
adiu
s (m
m)
Rot
or I
nner
R
adiu
s (m
m)
Rot
or O
uter
R
adiu
s (m
m)
Stat
or P
ole
Ang
le (
Deg
)
Rot
or P
ole
Ang
le (
Deg
)
Stat
or I
nner
R
adiu
s (m
m)
Tor
que
Ave
RP
SO(N
.M)
Tor
que
Rip
ple
Dev
iati
on
[18] 10 21 32 35.39 30.36 53 0.74 0.24 -
RPSO 4 22 38 32 45 58 0.84 0.17 -
0.85 0.77 5%
0.14 0.15 12%
rand * ±1 mm + RPSO Parameters
rand * ±1 mm + [18] ParametersRan
dom
Nei
ghbo
r
Wor
st
Figure 6 shows the torque profile and geometry of SRM design for
proposed RPSO and methodof [18] during geometry optimization. As
seen from Table 1 and Fig. 6, the improvement made by theproposed
method in the average torque value and ripple factor for optimized
structure are, respectively,13% and 30% more than those of previous
works in [18]. This shows the effectiveness of the proposedmethod
for the optimization of the SRM structure.
(a) (b)
Figure 6. Design comparison of RPSO and [18] in geometry
optimization: (a) Torque profile, (b)Geometry of SRM.
-
Progress In Electromagnetics Research M, Vol. 45, 2016 203
(a) (b)
Figure 7. Robustness comparison in geometry optimization: (a)
RPSO design and 4 random neighbors,(b) [18] design and 4 random
neighbors.
The results of the proposed method are also robust, which means
that the objectives functions willnot considerably change due to
defined tolerate of their parameters. Such as geometry variables
shownin Fig. 1 might change a bit due to the errors and
non-accuracy of the industrial machine using forproduction. The
acceptable tolerance of geometry variables is 1 mm. The accuracy of
a CNC machinedoes not exceed 0.1 mm. Considering the effect of
aging and low accuracy of a man-made motor usingan angle grinder or
other metal cutters, the dimensional tolerance is not more than
1mm. In Fig. 7, themain optimized geometry with RPSO comes with 4
random samples in a 2-D space, which representsthe average torque
and torque ripple in each axis where parameters of the optimized
SRM geometry arerandomly changed by 1mm. Table 1 shows the worse
results of random samples and objective functionvalues deviation.
As seen, the worst deviation of objective functions for RPSO
solution is better than[18] that shows the proposed method has
better robustness.
4.2. SRM Drive Robust Design
This research also contains minimization of the fitness function
(7) by optimizing the drive current valuein each two-degree
rotation of the rotor where the geometry parameters are according
to the results ofSeeker Optimization Algorithm [18].
The optimized current profile of SRM drive is expressed in Table
2. Fig. 8(a) shows torque profilecomparison of RPSO and [18] for
drive optimization where the RPSO optimized currents profile
isillustrated in Fig. 8(b).
As seen from Table 2 and Fig. 8(a), the reduction made by the
proposed method in the ripple factorfor optimized current profile
is 34% with respect to previous work [18]. This shows the
effectiveness ofthe proposed method for the optimization of the SRM
structure.
A Texas Instrument SRM drive is designed to have accuracy about
0.25%. As our SRM uses2.5 A current supplier connected to the
stator windings, therefore the accuracy is about 6 mA. Unlikethe
RPSO model, there is no robustness in the suggested optimized model
of [18] as shown in Table2. In Fig. 9, the optimized drive with
RPSO comes with 4 random samples in a 2-D space, whichrepresents
the average torque and torque ripple in each axis where the
parameter of optimized SRMdrive is randomly changed by 6 mA. Table
2 shows the worst results of random samples and objectivefunction
values deviation. As seen, the worst deviation of objective
functions for RPSO solution (4%)is better than [18] (24%) that
shows the proposed method has better robustness.
The results show that there is no considerable deviation from
the RPSO solution by randomlychanging the variables by 1mm for
geometry and 6mA for driver current. As seen, the RPSO solution
-
204 Ketabi, Yadghar, and Navardi
(a)
(b)
Figure 8. Drive optimization: (a) Torque profile comparison of
RPSO and [18], (b) RPSO optimizedcurrents profile.
Table 2. Current value of optimised result and the worst random
neighbor and the deviations of worstneighbor from the main
result.
Dri
ve
Mod
el
Tor
que
Ave
(N
.M)
Tor
que
Rip
ple
Sc
alar
D
evia
tion
fro
m
RP
SO
mA mA mA mA mA mA mA mA mA mA mA mA mA mA mA
2500
2500
2500
2500
2500
2500
2500
2500
2500
2500
2500
2500
2500
2500
2500 0.74 0.24 -
2500
2495
2497
2500
2500
2500
2500
2497
2495
2492
2490
2500
2500
2500
2500 0.75 0.16 -
W
orst
Ran
dom
Nei
ghbo
r 0.76 0.20 4%
0.75 0.43 24%
1o 3o 5o 7o 9o 11o 13o 15o 17o 19o 21o 23o 25o 27o 29o
rand * ±6 mA + RPSO Parameters
rand * ±6 mA + [18] Parameters
[18]
RPSO
-
Progress In Electromagnetics Research M, Vol. 45, 2016 205
(a) (b)
Figure 9. Robustness comparison in drive optimization: (a) RPSO
current profile design and 4 randomneighbors, (b) constant current
profile ([18]) design and 4 random neighbors.
is better in both torque and ripple than [18]. Also the random
neighbors do not exceed more than only5% from RPSO solution, then
their robustness is provable.
5. CONCLUSION
In this article, the RPSO algorithm has been employed to solve
SRM geometry and drive robustoptimization problems. A MATLAB linked
Finite Element Software, COMSOL 4.2, has been usedto torque
calculation and objective functions evaluation.
The proposed robust optimization method was verified on two case
studies for geometry and drivedesign. The results show an increase
in the torque value, a decrease in torque ripple and an increase
inrobustness compared with the previous optimized SRM. As a proof
for the robustness of the algorithm,the selected RPSO results have
been compared to its neighbors. The deviation from the main
modelwas renounceable, indicating the practicality for
manufacturing of the selected geometry and drive.Therefore, robust
optimization is necessary for the modern shape and drive design of
all types ofelectric motors.
The future work, which will be investigated, is the
multiobjective approach for the SRM designoptimization. Moreover,
material model will be investigated for the motor design
optimization.Thereafter, an experimental study will be presented
with details to verify the efficiency of these proposeddesign
optimization methods.
REFERENCES
1. Arkadan, A. A., H. H. Shehadeh, R. H. Brown, and N. A. O.
Demerdash, “Effects of choppingon core losses and inductance
profiles of SRM drives,” IEEE Transactions on Magnetics, Vol.
33,No. 2, 2105–2108, 1997.
2. Vujicic, V. P., “Minimization of torque ripple and copper
losses in switched reluctance drive,” IEEETransactions on Power
Electronics, Vol. 27, No. 1 388–399, 2012.
3. Fort, J., B. Skala, and V. Kus, “The torque ripple reduction
at the drive with theswitched reluctance motor,” 15th International
Power Electronics and Motion Control Conference(EPE/PEMC),
DS2a.16-1–DS2a.16-4, IEEE, 2015.
4. Niapour, S. A., K. H. Mozaffari, M. Tabarraie, and M. R.
Feyzi, “A new robust speed-sensorlesscontrol strategy for
high-performance brushless DC motor drives with reduced torque
ripple,”Control Engineering Practice, Vol. 24, 42–54, 2014.
-
206 Ketabi, Yadghar, and Navardi
5. Su, G.-W., M.-Y. Cheng, and W.-C. Chi, “Current loop
controller design for torque ripplesuppression of switched
reluctance motors,” 2013 CACS International Automatic
ControlConference (CACS), 496–500, IEEE, 2013.
6. Nakazawa, Y., K. Ohyama, H. Fujii, H. Uehara, and Y.
Hyakutake, “Improvement of efficiency ofswitched reluctance motor
by single pulse control based on linear torque equation,” 15th
EuropeanConference on Power Electronics and Applications (EPE),
1–10, IEEE, 2013.
7. Singh, S. K. and R. K. Tripathi, “Minimization of torque
ripples in SRM drive using DITC forelectrical vehicle application,”
2013 Students Conference on Engineering and Systems (SCES),
1–5,IEEE, 2013.
8. Kalaivani, L., P. Subburaj, and M. W. Iruthayarajan, “Speed
control of switched reluctancemotor with torque ripple reduction
using non-dominated sorting genetic algorithm
(NSGA-II),”International Journal of Electrical Power & Energy
Systems, Vol. 53, 69–77, 2013.
9. Mikail, R., I. Husain, Y. Sozer, M. S. Islam, and T.
Sebastian, “Torque-ripple minimizationof switched reluctance
machines through current profiling,” IEEE Transactions on
IndustryApplications, Vol. 49, No. 3, 1258–1267, 2013.
10. Dowlatshahi, M., S. M. Saghaeian Nejad, and J. Ahn, “Torque
ripple minimization of switchedreluctance motor using modified
torque sharing function,” 2013 21st Iranian Conference onElectrical
Engineering (ICEE), 1–6, IEEE, 2013.
11. Ma, C., L. Qu, and Z. Tang, “Torque ripple reduction for
mutually coupled switched reluctancemotor by bipolar excitations,”
2013 IEEE International Electric Machines & Drives
Conference(IEMDC), 1211–1217, IEEE, 2013.
12. Somesan, L.-E., E. Padurariu, and I.-A. Viorel, “Two simple
analytical models, direct and inverse,for switched reluctance
motors,” Progress In Electromagnetics Research M, Vol. 29, 279–291,
2013.
13. Bist, V. and B. Singh, “A brushless DC motor drive with
power factor correction using isolatedzeta converter,” IEEE
Transactions on Industrial Informatics, Vol. 10, No. 4, 2064–2072,
2014.
14. Jin, Y., “Advanced control methods for torque ripple
reduction and performance improvement inswitched reluctance motor
drives,” PhD diss., 2014.
15. Youssef, M. Z., “Design and performance of a cost-effective
BLDC drive for water pumpapplication,” IEEE Transactions on
Industrial Electronics, Vol. 62, No. 5, 3277–3284, 2015.
16. Parackal, R. and R. A. Koshy, “PV powered zeta converter fed
BLDC drive,” 2014 AnnualInternational Conference on Emerging
Research Areas: Magnetics, Machines and Drives(AICERA/iCMMD), 1–5,
IEEE, 2014.
17. Abraham, A. and A. Mathew, “Implementation of a novel PFC
Cuk rectifier fed brushless DCmotor drive,” 2014 Annual
International Conference on Emerging Research Areas:
Magnetics,Machines and Drives (AICERA/iCMMD), 1–5, IEEE, 2014.
18. Navardi, M. J., B. Babaghorbani, and A. Ketabi, “Efficiency
improvement and torque rippleminimization of switched reluctance
motor using FEM and seeker optimization algorithm,”
EnergyConversion and Management, Vol. 78, 237–244, 2014.
19. Xia, C., Y. Xiao, W. Chen, and T. Shi, “Torque ripple
reduction in brushless DC drives basedon reference current
optimization using integral variable structure control,” IEEE
Transactions onIndustrial Electronics, Vol. 61, No. 2, 738–752,
2014.
20. Dowlatshahi, M., S. M. Saghaian Nejad, M. Moallem, and J.
Ahn, “Torque ripple reductionof switched reluctance motors
considering copper loss minimization,” 2014 IEEE
InternationalConference on Industrial Technology (ICIT), 858–865,
IEEE, 2014.
21. Buja, G., M. Bertoluzzo, and R. K. Keshri, “Torque
ripple-free operation of PM BLDC drives withpetal-wave current
supply,” IEEE Transactions on Industrial Electronics, Vol. 62, No.
7, 4034–4043,2015.
22. Gao, X., X. Wang, Z. Li, and Y. Zhou, “A review of torque
ripple control strategies of switchedreluctance motor,”
International Journal of Control & Automation, Vol. 8, No. 4,
103–116, 2015.
23. Balaji, M. and V. Kamaraj, “Evolutionary computation based
multi-objective pole shapeoptimization of switched reluctance
machine,” International Journal of Electrical Power &
Energy
-
Progress In Electromagnetics Research M, Vol. 45, 2016 207
Systems, Vol. 43, No. 1, 63–69, 2012.24. Tavakkoli, M. A. and M.
Moallem, “Torque ripple mitigation of double stator switched
reluctance
motor (DSSRM) using a novel rotor shape optimization,” 2012 IEEE
Energy Conversion Congressand Exposition (ECCE), 848–852, IEEE,
2012.
25. Ren, Z., D. Zhang, and C. S. Koh, “Multi-objective
worst-case scenario robust optimal design ofswitched reluctance
motor incorporated with FEM and Kriging,” 2013 International
Conferenceon Electrical Machines and Systems (ICEMS), 716–719,
IEEE, 2013.
26. Chiariello, A. G., A. Formisano, R. Martone, and F. Pizzo,
“Gradient-based worst case searchalgorithm for robust
optimization,” IEEE Transactions on Magnetics, Vol. 51, No. 3,
7205004,2015.
27. Hong, W. C., “Chaotic particle swarm optimization algorithm
in a support vector regression electricload forecasting model,”
Energy Conversion and Management, Vol. 50, No. 1, 105–117,
2009.
28. Xiao, S., Y. Li, M. Rotaru, and J. K. Sykulski, “Six sigma
quality approach to robust optimization,”IEEE Transactions on
Magnetics, Vol. 51, No. 3, 7201304, 2014.
29. Mirjalili, S., A. Lewis, and S. Mostaghim, “Confidence
measure: A novel metric for robust meta-heuristic optimisation
algorithms,” Information Sciences, Vol. 317, 114–142, 2015.
30. Lei, G., T. Wang, J. Zhu, Y. Guo, and S. Wang, “System level
design optimization method forelectrical drive systems — Robust
approach,” IEEE Transactions on Industrial Electronics, Vol. 62,No.
8, 4702–4713, 2015.
31. Malarvizhi, K. and M. Kumar, “Particle swarm optimization
tuned BELBIC controller for 8/6SRM operation,” 2015 2nd
International Conference on Electronics and Communication
Systems(ICECS), 904–909, IEEE, 2015.
32. Angeline, P. J., “Evolutionary optimization versus particle
swarm optimization: Philosophy andperformance differences,”
Evolutionary Programming VII, 601–610, Springer, Berlin,
Heidelberg,1998.
33. Li, M., R. Silva, F. Guimarães, and D. Lowther, “A new
robust dominance criterion formultiobjective optimization,” IEEE
Transactions on Magnetics, Vol. 51, No. 3, 8201504, 2015.
34. Mirjalili, S. and A. Lewis, “Novel performance metrics for
robust multi-objective optimizationalgorithms,” Swarm and
Evolutionary Computation, Vol. 21, 1–23, 2015.